sghai9/technology-mediation-csea

This contains the full analysis workflow for the study on technology-facilitated child sexual exploitation and abuse (CSEA) in Eastern and Southern Africa and Southeast Asia. ## Overview This repository contains the analytical code for a study examining technology-facilitated child sexual exploitation and abuse (CSEA) across Eastern and Southern Africa and Southeast Asia. The analysis uses data from the Disrupting Harm study conducted by UNICEF, ECPAT International, and INTERPOL. ## Repository Structure ``` project/ ├── reference-data/ # ITU data for population burden validation ├── qmd/ # Analysis scripts ├── 1a_preprocess.html # Rendered outputs ├── 1b_population_burden.html ├── 2a_disclosure.html ├── 2b_CSEAmodelling.html ├── 3a_disclosure_modelling.html ├── 3b_disclosure_modelling_imp.html ├── LICENSE └── README.md ``` ## System Requirements ### Software dependencies | Software | Version tested | |---|---| | R | 4.4.1 (2024-06-14) | | CmdStan | 2.32.2 | | Quarto | 1.4.x | Required R packages: ```r install.packages(c( "tidyverse","rlang","survey","srvyr","mice","brms","cmdstanr","loo","emmeans", "bayesplot","bayestestR","posterior","marginaleffects","ggdist","ggrepel", "ggnewscale","ggh4x","viridis","RColorBrewer","colorspace","scales","corrplot", "knitr","kableExtra","gt","gtsummary","broom.mixed","sjPlot","sf", "rnaturalearth","rnaturalearthdata","DataExplorer","patchwork","gridExtra", "cowplot","naniar","collapse","flextable","officer","png","readxl","ggtext", "glue","readr","stringr","gridGraphics","janitor","here", "rstan","car","performance","tidybayes" )) # CmdStan must be installed separately: cmdstanr::install_cmdstan() ``` ### Non-standard hardware No non-standard hardware was required. The modelling scripts use parallel computation with up to 8 CPU cores. ## Rendered Outputs > **A demo dataset is not available.** The data used in this study involve sensitive child-level responses collected under a restricted access agreement with UNICEF and cannot be shared publicly. We considered generating synthetic data but determined this could risk misuse or misinterpretation given the sensit …